Servers by jDrones

This should be interesting


(Bill Geyer) #701

Great news. So you are saying that with the notch the instability is showing up at 4 hz. I see no issue with using some D gain to see if that helps with the pitch bounce. If you get the D gain up high enough you may see the instability start to crop up on the high side of the notch filter.


(Timothy Bacon) #702

Bill,
Without notch the instability is 5 and 8hz pitch & roll. Once I add in the notch, it drives it down to 4, likely at the other filter’s knee? In any case the amplitude is substantially reduced once the notch is in place, so instead of .04 P term, i can get all the way up to almost .2 before real P term ossicilations pop up. I settled at .16 for now to keep the pitch bounce down, but if D term helps with that i can likely run about .18 safely and I term in the same neighborhood. If finding out that it feels much more stable and in control when I term and P term are somewhat proportional. When ilI had low P term and high I term it just felt mushy, and VFF just seemed to add in a snappy quality to it, not a smooth control feel.
In any case, I am quite happy with where I left it parameter wise, so far it by far the best its flown. I still have a ways to go fine tuning and working on the pitch bounce. A quick loiter test went swimmingly as well.
Tim


(Bill Geyer) #703

Maybe I didn’t catch this but is the FILT parameter still at 20 hz?


(Timothy Bacon) #704

Bill,
The FILT parameter is at 4hz.


(Pitt) #705

Have you tried FILT = 4Hz without notch filter?
Just wanted to know

The output notch filter I tried caused lower frequency oscillation without increasing P gain.
I’ll be trying the new input filter soon.


(Bill Geyer) #706

Pitt,
Yes, Tim had tried using FILT =4 with no notch filter and I believe he settled on roughly 0.05 as the best P gain for that configuration. [quote=“PittRBM, post:705, topic:16800”]
I’ll be trying the new input filter soon.
[/quote]
Is this the filter I provided you that notches only the gyro signal? Or did you find something else? Just curious.


(Pitt) #707

It’s the one integrated in 3.5.
Since I have to complie the code myself anyway, I also added your adjustable I_LEAK commit to my 3.5 firmware but I don’t feel like moving yet.


(Bill Geyer) #708

So it isn’t the one that I provided? What’s the param that you use to adjust it? Is it a notch filter?


(Pitt) #709

The one I have tried is the one you provided which was an output filter (which you recently pointed me to a fix to gyro signal filter).

Your new gyro signal filter compiled but not used yet.


(Timothy Bacon) #710

Pitt,
Confirming what Bill had said above already, yes I had tried the filter at 4hz. It helped me get the P term above .02-.03 without really bad tendencies, but ultimately I could not go above .05 before the ossicilations piled up.
Basically I am finding that filter choice, whether it is the in-built low pass filter or the notch bill just helped me implement, determine where the instability will pop up. I notched at X frequency to attenuate the problem areas of the spectrum i detemined by setting the lowpass at 20 hz and seeing where they started. Then i notched the frequencies for pitch and roll as well as setting the lowpass at 4hz and have acheived great resluts. There is now a very weak instability in the 3-4hz region, but it is nothing in comparison to the issues before. So using the filter seems to have moved an instability to a lower frequency. That being said the positive effects of the filter without any rigorus tuning have been nothing short of stellar. It is truly an amazing difference in the handling quality. I immediately felt “at home” on the sticks to coin a phrase and proceeded to fly it like I stole it. :slight_smile:
I would be completely content leaving it where I did yesterday but the reality of it is that I started from scratch and only put a few flights on it and nailed it down 100% better than it was before, so im guessing with a closer look at the data and some more tuning flights it can only get better.
My current settings: Low pass filter 4hz both PIT and RLL
Notch filter 5hz PIT 7hz RLL
P term .154 RLL. .13 PIT
I term .18 PIT. .19 RLL
D term .001 PIT. .001 RLL
ILMI PIT .07 ILMI RLL .07
VFF PIT .06 VFF RLL .05
I LEAK RATE .02
IMAX PIT .4 IMAX RLL .4

With these settings if feels very much like a standard FBL unit. I think that is why I am happy with the results so far. I am running lower P term on the PIT axis which is counter intuitive from the norm I guess as from what I have read the RLL axis is usually the first to have issues, but since I have a bit of a bounce I dialed it back for the short term.

I believe Chris has hit the nail on the head in that it is a product of the headspeed I am running. That being said, I am going to try to feather in some D term and see if that helps the bounce a bit? Worse comes to worse I will bump up the headspeed a bit.
I just know that before the filter, adding more headspeed required me to keep lowering the P term from the already low value it was at so I split the difference per se’ before.
I am also going to see what happens when the Low pass filter is set to something like 10hz as well. Basically at the moment I am putting aside the end result I have in mind for the sake of data collection. This produced results that are incredibly promising I feel for the traditional helicopter side of Copter, I always have the Octa Copter for work if I need to do a job so the heli is going to be a test platform for the near future now. Turning out to be very good stuff here! :slight_smile:
Tim


(Chris Olson) #711

Tim, it normally is a problem with low headspeed. And is the first thing I would suspect. But not always. In helicopters it can also be a CG issue.

I would recommend caution with your current settings in high speed auto flight. I have been the route you are on right now and did not have good results with it when I started pushing the speeds up with the autopilot flying the helicopter. I never engaged the notch filter because my helicopter did not require it. I got mine to .19 on P gain without the notch filter. But I ran into a serious pitch instability in FF flight with the autopilot starting at about 15m/s flight speed on a helicopter that was otherwise fine with me flying it at 20m/s. What set it off was coming out of a spline turn, crossing the waypoint and getting the “jerk” or “twitch” that you get mixing standard with spline waypoints. The autopilot lost control of the helicopter and I got it back by switching to Stabilize and going to full nose up at full collective to brake it. It did pretty much the same thing I’ve seen FBL controllers do trying to fly a heli at high speed with them with the gains set too high.

So just be careful with it and don’t fly your first auto flights at much more than the default speed until you get a chance to see how bad those “twitches” are when crossing waypoints.


(Timothy Bacon) #712

So I managed to get a couple flights in before it started raining bad earlier.
I left everything as it was and upped the D term to .007 from .001. Unfortunately i had an ossicilation that built quickly so I had to set it down quick. I lowered the D term to .005 amd flew again, just a hint of ossicilation so I landed and lowered to .003 and settled there.
Just that extra little bit of D term seems to help stability and it did indeed get rid of some of the bounce I am seeing with the pitch axis. Just for the sake of testing and data collection I am going to lower the notch filter on pitch axis to 4.5hz and see if it changes anything. If not I will set notch back to 5hz on pitch and leave D term where it is as it was flying great save for the pitch bounce.
Much as i hate to as it will reduce flight time, i will raise headspeed if I have to per Chris’s sound advice and check tuning on all parameters and reduce the gains across the board if nessecary.
Will be able to report back with my findings on reduction of the notch filter frequency later provided the rain will quit. Unfortunately, as it sits currently I dont think I will be able to raise D term much higher to get a reasonable contribution to the flight characteristics. Time will tell.

I will say it does great in forward flight with the P term where it is now, it just tracks far better across the board. I ran a quick backwards circuit moving at a good clip and it looked far better than it did before the new filter.
Tim


(Timothy Bacon) #713

Chris,
I appreciate the advice! I am definitely going to take this one step at a time and alowly build up the speeds with auto flight. I will keep distances low from my location so if need be i can flip into stabilize and take it back. Thats the plan anyways… :confused:
And as i noted above, i was really hoping to solve the bounce without adding in more power robbing headspeed, but I just don’t think the filtering and tunability is there quite yet in Copter to solve it, so 50-100rpm is likely the best medicine.
Tim


(Timothy Bacon) #714

On another note while im thinking about it, without question, I have found that the closet I get P I D terms closer to balanced, the helicopter just seems to feel better and more natural in flight.
When I had low P, high I and higher VFF, it just did not feel right. Now that im able to run P a bit higher, I a bit lower and even just ablittle bit of D with a small reduction of VFF, it feels altogether different and less foriegn to me. The change is quite evident in watching it fly as well. My better half even noticed it looked alot better in the air and she is certainly not a pilot but has watched me fly quite a bit. To quote her words, she said it dident look like I was chasing it anymore.
Tim


(Chris Olson) #715

Helicopters need considerable I gain, and less P in FF flight. Much more on I than they need in hover. They just fly different than multi’s. The software includes a rudimentary method (the I term “leak”) to adjust for the difference in hover vs FF. But it’s not as smooth as flybar.

I’ve done it with as little as 50 rpm change. Sometimes it doesn’t take much to get that stability in the pitch axis. I have always gone by hover collective. If it won’t hover on 5 degrees you are likely running the head too slow. If it hovers on 5, then it’s another issue like maybe CG, which can still be solved by running the head faster.

Your blades (symmetrical airfoil) are probably most efficient lift/drag at 4 degrees. In ETL that’s what they should be running at. So if you slow the head down and have to use say 6-7 degrees to hover, you will likely not see any “power robbing” effect by running the head faster and getting the AoA of the blades to a more efficient range. The speed of the head is not the only factor in how much shaft horsepower it takes to turn the rotor at a particular loading. The blades have an efficient range (best l/d ratio), creating lift, from 3-6 degrees AoA. And increasingly inefficient range from 6 to 14.


(Bill Geyer) #716

So you may not have needed the notch in a hover but forward flight can destabilize the rotor especially in AoA. So you probably did need a notch to help with forward flight. In looking at one of your flights from the DFC, I did notice that there was some small oscillatory stuff going on which you may not have seen in a hover but may have gotten worse with forward flight. So the autopilot could have set that off. Now I’m not saying that with the high P gain Tim isn’t going to have some pitch oscillation problem in forward flight, but the aircraft is better protected with the notch. I have seen in multicopters that in forward flight in auto mode, there is a bad pitch oscillation but much lower frequency like 1 hz.
So I agree with your caution Chris. Tim needs to use good build up and review data to ensure there are no instabilities being excited, especially this pitch bob.

I had changed head speed with the X-3 to look at impact on instability freq and I don’t recall there being much of an impact escpecially with 50RPM difference. Let me know what you find out.
Also I think you need to lower the leak rate of the integrator (I_LEAK_RATE= 0.001), this will allow more I term to build if error is present and you will notice that you may need less I gain in a hover. Now the I leak rate parameter will only affect hover as it is turned off in forward flight.[quote=“ChrisOlson, post:715, topic:16800”]
Helicopters need considerable I gain, and less P in FF flight. Much more on I than they need in hover.
[/quote]
I don’t know about this. Certainly the integrator will need to be able to retain more error because of the rotor blowback in forward flight and additional longitudinal stick is necessary to keep the tip path plane forward. So you will need to look at IMAX to make sure that the integrator is not being limited in high speed forward flight. I’m not sure about the I gain necessarily needing to be higher. I think we need to look at the data and see what’s going on.


(Chris Olson) #717

I’m not sure if it was the flight dynamics combined with the helicopter having to make a fast move at the waypoints, or what. I only know it became highly unstable. I was running 1,950 rpm headspeed on the 500 then, and I have since ramped that up to 2,100. But the additional headspeed didn’t help because it showed up again at few m/s faster speed.

I consider the pitch bob to be a dangerous issue for high speed flight. I would not even attempt it until it’s fixed. If it starts a rapid porpoising in FF flight you will lose control of the helicopter.


(Timothy Bacon) #718

Chris, quickly I will say I have managed to reduce the pitch bob immensely. I tried the headspeed bump, even went up 250rpm and the problem was still there. I could see the stability the headspeed increase gave, but still dident fix the issue.
So I went back to tuning, changed I gain, P gain and moved the notch filter. 5hz and its all but gone. No pitch bob.
New issue though. If I flip into loiter quickly without making sure the helicopter is stationary first, i get a slow toilet bowl. It will settle on its own, but takes a couple rotations to do so.
I am finding that I solve one problem and create another. Although the major ossicilations are gone and flight performance feels light years better as a whole.
Ill have to sift through the logs tonight and see how its tracking.
I guess this is just a case of charge, fly, tweak and repeat until its right…
Honestly I think picking a parameter and adjusting to see the effects and then setting it at the best value then moving on to another is helping me understand the effects each one has.

I have a good base now with which to work with and just need to navigate the waters and find what works best for this frame and how I want it to fly.
Tim


(Bill Geyer) #719

I[quote=“timbaconheli, post:718, topic:16800”]
New issue though. If I flip into loiter quickly without making sure the helicopter is stationary first, i get a slow toilet bowl. It will settle on its own, but takes a couple rotations to do so.
[/quote]
Chris back me up on this. I seem to remember Rob thinking this was an issue with copter 3.5.


(Chris Olson) #720

Yes, this is correct. And I have not had a chance to go back to 3.5 to test it further. I attributed it to possibly having to rescale my accels for Loiter at the time I tested it. Rob has not come up with a solution or cause yet. Randy said there is nothing different with the Loiter controller in 3.5. So we don’t know.